Merge checklist

Steps for merging refactor/modular into main

⚠️ HISTORICAL. This merge happened.

Status added 2026-08-12. origin/refactor/modular still EXISTS as a ref but is no longer used for anything — main is the only live branch, and the maintainer commits to it directly. Do not restart the two-branch flow because the ref is still there. The "only system in active campus use" line below is also out of date: SAMO Shop, ระบบบ้าน, ทีม SAMO and หนังสือโครงการ are all live now.

Kept because the data-safety sequencing is a good model for the next big cutover. Not a checklist anyone should be working through today.

Goal: ship the refactor to production without losing or contradicting PR data, which is the only system in active campus use right now. Announcements and VitalSound have low-to-zero live volume.

TL;DR — safe path

The refactor branch is mostly backwards-compatible at the data schema level. The risky things are:

1. Staff credentials changed (frontend + backend).
2. GAS URLs in config.js point at DEV deployments, not prod.
3. Announcements sheet got a new column 7 (thumbnail).

Don't merge until you've handled all three.

---

Step 0 — Snapshot prod data

Before doing anything, take a copy:

• In the PR Submissions Google Sheet: File → Make a copy → rename Submissions backup YYYY-MM-DD.
• Same for Announcements sheet and the VS Tickets sheet.

This is your rollback. Don't skip this.

Step 1 — Decide on staff credentials in prod

The refactor uses samomdkkupr / samomdkkuvssound / samomdkkudev. Prod GAS has prsamomdkku (PR) and samomdkku69 (VS) — different.

Options:

• (A) Update prod GAS to match refactor (recommended). After this, the existing staff need to be told the new login is samomdkkupr + «disabled 2026-08-17».
• (B) Update refactor frontend to match prod creds, then merge. Set STAFF_ACCOUNTS in src/js/auth.js to the prod usernames; also update handleVerifyPRStaffLogin in appscript/prform.gs and verifyStaffLogin in appscript/vssound.gs to match.

Whichever you pick, the frontend STAFF_ACCOUNTS keys and the backend hardcoded strings must agree — otherwise staff can't log in.

Step 2 — Point config.js at prod GAS

⛔ THIS STEP NO LONGER APPLIES — and both URLs it printed were wrong by 2026-09-10. It described config.js as holding the DEV /exec URLs with prod in a comment, which stopped being true long ago: the file exports one live GAS_API_URL and nothing else. Checked 2026-09-10:

• the GAS_API_URL written here (AKfycbw1iHE4ALCO…) is not the endpoint the app calls (AKfycbwomKii…), and
• GAS_VITAL_SOUND_URL does not exist — vssound.gs was deleted when Discord moved to the /notify service, and the name has 0 references in src/.

There is nothing to switch, so do not switch anything. Read the live value from its one home when you need it:

grep -A2 'export const GAS_API_URL' src/js/config.js

Kept rather than deleted because a checklist step that vanishes looks like one somebody forgot. A copied URL is a decaying fact with a second home, and this is the third place that bit: skills/deploy-gas.md printed the same stale URL.

Better long-term: wire import.meta.env.VITE_GAS_API_URL with the production URLs configured per-environment in the Cloudflare Pages dashboard, then this file never needs editing again. The 30-line change is in config.js — say the word and it gets done.

Step 3 — Deploy backend changes to prod GAS

Open appscript/prform.gs and appscript/vssound.gs in this repo, copy each function that changed into the prod Apps Script editor (not dev), then Deploy → Manage deployments → Edit → New version. The prod /exec URLs stay the same.

Changes that need to land in prod GAS:

prform.gs

• handleVerifyPRStaffLogin — staff cred per Step 1.
• handleAddAnnouncement + handleEditAnnouncement + doGet getAnnouncements — thumbnail column 7. Backwards-compatible: legacy rows just have no thumbnail; new rows write to col 7.
• handlePRSubmission — gated Discord call (data.skipDiscord short- circuits). Backwards-compatible: prod gets the gate but old clients never set the field, so behavior is identical.

vssound.gs

• verifyStaffLogin — staff cred per Step 1.
• handleVitalSoundSubmit — gated Discord call. Same backwards-compat story as PR.
• getUserHistory — trustClient flag changes the empty-result wording but the success path is unchanged. Backwards-compatible.

Announcements sheet

Add a header for column G manually: Thumbnail. The GAS will write to G regardless, but a labeled header keeps the data readable.

Step 4 — Verify PR data won't contradict

The refactor's PR submit path writes the same columns to the same sheet schema. New fields, all in existing columns:

• submitterEmail (col 18) — previously held Google emails or "Guest". Now also holds @<username> strings for password users. Old rows unaffected; new rows just have a different identifier format. No contradiction; the value is treated as opaque.
• Status/dates/everything else: identical.

PR staff dashboard (read side): the kanban renderer buckets tickets by status. Statuses not in the canonical list bucket into "รอ PR รับเรื่อง" via the substring fallback in pr-staff.js. Old prod tickets with arbitrary status strings (if any) won't break — they get a column.

Conclusion: PR data is safe. Old tickets readable, new tickets written in compatible format.

Step 5 — Verify announcement data

Refactor reads post.thumbnail and falls back to first content image (existing behavior) when missing. Old announcements with no col 7 render unchanged.

Conclusion: Announcement data is safe.

Step 6 — Verify VS data

The refactor's VS submit writes vsUsername + vsPassword for both auth methods. Old prod rows with explicit username/password still work in getUserHistory. Caveat: existing VS users who registered under prod's mode=create flow and never submitted a ticket won't appear; this was already true in prod.

Conclusion: VS data is safe; user experience for not-yet-submitted users isn't worse than today.

Step 7 — Test on Cloudflare Pages preview

Cloudflare auto-deploys every branch. Before merging:

1. Open the preview URL for refactor/modular (formed from the branch name in the Cloudflare dashboard).
2. With Step 2's prod URLs still pointing at dev GAS, smoke-test every flow: PR submit (then check the dev sheet), PR track, sign in as each staff role, kanban filter, announcement create/edit.
3. Optional rehearsal: temporarily set Cloudflare preview env vars to point at the prod GAS, smoke-test against real data, then flip back to dev. (No code changes needed.)

Step 8 — Merge

# Make sure config.js points at prod URLs (Step 2)
# Make sure prod GAS has the new code deployed (Step 3)
git checkout main
git pull origin main
git merge --no-ff refactor/modular -m "merge: refactor/modular into main"
git push origin main

Cloudflare Pages auto-deploys main to production within ~60s.

Step 9 — Watch for ~24h

• Check the PR Discord webhook is still firing for real submissions (the most visible failure mode).
• Spot-check the Submissions sheet a few times — does the new column format look right? Are timestamps correct?
• If a staff member can't log in, they're hitting the credentials change from Step 1 — communicate the new login.

Rollback

If the merge breaks something critical:

git checkout main
git revert -m 1 <merge-commit-sha>
git push origin main

Cloudflare re-deploys the previous version. GAS changes don't auto- roll back — manually revert the Apps Script deployments to the prior version via Deploy → Manage deployments → Archive newest.

What this checklist does NOT cover

• Migrating to Supabase. See SUPABASE-MIGRATION.md.
• Unifying the user data model. See AUTH-MODEL.md.
• Adding RLS / proper auth. Both above.